Research Summary: The Tethys Dataset: Seven Years of Hourly Smart Water Metering and a Pipeline for Making It Usable

A new research publication introduces the Tethys dataset, comprising 91 months of hourly water consumption data from 24 municipal buildings. This dataset is notable for its explicit quantitative account of data quality, addressing common issues with existing water consumption datasets that often lack transparency regarding data cleaning processes and deployment defects. Key findings reveal significant data gaps, including fleet-wide outages, and identify an aggregation error in the released files that introduced over 200,000 impossible cumulative index decreases.

Why it matters

This research is strategically important because it highlights critical challenges in data quality, availability, and processing that underpin foundational methods for water demand forecasting and leak detection. Addressing these issues can significantly improve the accuracy and reliability of water management systems, informing better infrastructure investment and resource allocation decisions.

Key insights

• The Tethys dataset provides 91 months of hourly water consumption from 24 municipal buildings, published with a quantitative account of its quality.
• Raw data availability stands at 59.4%, indicating substantial gaps.
• Data loss is not random, with four fleet-wide outages totaling 595 days impacting the entire dataset simultaneously.
• Aggregation processes used to generate the released files silently introduced 205,200 impossible decreases in the cumulative index.
• The identified aggregation error is correctable with a one-line code change.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2609.22358